{"id":51534453,"url":"https://github.com/janteichertkluge/synth-bench","last_synced_at":"2026-07-09T07:01:08.385Z","repository":{"id":356651910,"uuid":"1220051784","full_name":"JanTeichertKluge/synth-bench","owner":"JanTeichertKluge","description":"synthbench is a Python library for generating reproducible, metadata-rich synthetic datasets for benchmarking.","archived":false,"fork":false,"pushed_at":"2026-05-06T08:58:09.000Z","size":4758,"stargazers_count":2,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-05-09T05:16:39.466Z","etag":null,"topics":["benchmark","benchmarking","classification","data","datasets","dgp","evaluation","regression","synthetic-data"],"latest_commit_sha":null,"homepage":"https://janteichertkluge.github.io/synth-bench/","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/JanTeichertKluge.png","metadata":{"files":{"readme":"README.md","changelog":"CHANGELOG.md","contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-04-24T13:48:29.000Z","updated_at":"2026-05-06T08:57:17.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/JanTeichertKluge/synth-bench","commit_stats":null,"previous_names":["janteichertkluge/synth-bench"],"tags_count":2,"template":false,"template_full_name":null,"purl":"pkg:github/JanTeichertKluge/synth-bench","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/JanTeichertKluge%2Fsynth-bench","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/JanTeichertKluge%2Fsynth-bench/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/JanTeichertKluge%2Fsynth-bench/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/JanTeichertKluge%2Fsynth-bench/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/JanTeichertKluge","download_url":"https://codeload.github.com/JanTeichertKluge/synth-bench/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/JanTeichertKluge%2Fsynth-bench/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":35290235,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-07-09T02:00:07.329Z","response_time":57,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["benchmark","benchmarking","classification","data","datasets","dgp","evaluation","regression","synthetic-data"],"created_at":"2026-07-09T07:01:05.924Z","updated_at":"2026-07-09T07:01:08.378Z","avatar_url":"https://github.com/JanTeichertKluge.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"\u003cdiv align=\"center\"\u003e\n  \u003cimg src=\"icon.png\" alt=\"synthbench\" width=\"420\"\u003e\n\u003c/div\u003e\n\n---\n\nsynthbench is a small Python library for generating synthetic datasets that are actually useful for benchmarking. You control the signal complexity, add noise or missing data on top, and get back a dataset with full provenance so you know exactly what you generated and why. Every result is reproducible from a single integer seed.\n\nIt covers eight DGP families, five corruptors, metadata enrichment (Bayes error, effective rank), Parquet/CSV serialization, and sweep helpers for running ablation grids.\n\n## Installation\n\n```bash\npip install synthbench\n```\n\nFor Parquet support:\n\n```bash\npip install \"synthbench[io]\"\n```\n\nFor `RandomNeuralDGP` (needs PyTorch):\n\n```bash\npip install \"synthbench[neural]\"\n```\n\n## Basic usage\n\n```python\nfrom synthbench import BenchPipeline, LinearDGP, MissingDataCorruptor\n\npipeline = BenchPipeline(\n    LinearDGP(complexity=\"medium\", task_type=\"classification\"),\n    corruptors=[MissingDataCorruptor(proportion=0.1, mechanism=\"mar\")],\n)\nresult = pipeline.run(n_samples=500, n_features=10, random_state=42)\n\nprint(result.X.shape)                     # (500, 10)\nprint(result.metadata[\"bayes_error\"])     # empirical difficulty estimate\nprint(result.metadata[\"effective_rank\"])  # feature space dimensionality\n```\n\n## What it does\n\n**Data-generating processes** — Linear, Polynomial, Tree, Friedman (variants 1/2/3), Additive, Sparse, Geometric, and RandomNeural. Each takes a `complexity` parameter and records ground-truth feature importances alongside the data.\n\n**Corruptors** — MeasurementNoise, Outlier, MissingData, Collinearity, and Categorical corruptors for the feature matrix, plus `LabelNoiseCorruptor` for flipping labels or injecting regression noise. They chain together in a canonical order and track how much signal they degrade.\n\n**Metadata** — every result carries `bayes_error`, `effective_rank`, corruptor parameters, and version provenance. Enough to reconstruct the generating pipeline from scratch.\n\n**Sweeps** — `severity_sweep` and `difficulty_sweep` for single-axis ablations, and `experiment_grid` for full factorial runs across sample size, complexity, and severity. Seeds are derived hierarchically so cells are independent but deterministic.\n\n**Named suites** — `BenchSuite(\"easy-classification\").run()` returns a labelled dict of results for a curated collection. Good for quick sanity checks or as a shared benchmark baseline.\n\n**Serialization** — `to_parquet` / `from_parquet` and `to_csv` / `from_csv` round-trip everything including metadata. `BenchPipeline.from_metadata` reconstructs and re-runs the pipeline for bit-identical replay.\n\n## Ablation example\n\n```python\nfrom synthbench import LinearDGP, OutlierCorruptor, experiment_grid\n\ngrid = experiment_grid(\n    LinearDGP,\n    OutlierCorruptor,\n    n_samples_list=[200, 500, 1000],\n    complexities=[\"low\", \"medium\", \"high\"],\n    severities=[\"low\", \"medium\", \"high\"],\n    n_features=10,\n    random_state=0,\n    task_type=\"classification\",\n)\n\nresult = grid[(500, \"high\", \"medium\")]\nprint(result.metadata[\"bayes_error\"])\n```\n\n## Docs\n\nFull reference at [JanTeichertKluge.github.io/synth-bench](https://JanTeichertKluge.github.io/synth-bench).\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fjanteichertkluge%2Fsynth-bench","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fjanteichertkluge%2Fsynth-bench","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fjanteichertkluge%2Fsynth-bench/lists"}